← Cases
JapanMedia and entertainment2023 – ongoing

Nippon Television Network Corporation

FACTly-Mate generative AI chat and internal document search

Nippon Television reached about 2,000 unique browsers on its in-house generative AI chat within six months

WorkingVendor case studyVendor ↗

Problem

Staff at Nippon Television, which had 1,353 employees in April 2024, needed faster access to information in internal documents. Public generative AI services carried a risk of information leaks.

Approach

The project started in summer 2023 on top of the company's FACTly data platform, which is centered on BigQuery. The first phase gave all staff a secure generative AI chat environment with ready-made prompts for tasks such as contract translation and summaries. The second phase added document search with Vertex AI Search (RAG), with Gemini called before and after the search to classify queries. The model was first PaLM 2 and later moved to Gemini. Search covered a limited set of documents at the time.

Results

2,000前後unique browsers using the chat about six months after launch (May 2024), against about 1,400 employees, per Google Cloud's customer storyVendor case study[1]

Takeaway

A briefing session for 150 people and interviews with staff before launch were credited with a smooth rollout. Accuracy of document search needed extra prompt work and classification steps.

Categories

Innovation type
Process innovation
AI technology
Generative AI
Evidence
Vendor case study

Sources

Source summaries

[1]Google Cloud25 Jul 2024

Google Cloud's customer story describes Nippon Television's two-phase FACTly-Mate project, a secure generative AI chat followed by RAG document search. It reports about 2,000 unique browsers six months after launch and about 1,400 employees. The figures come from the company's own statements.

Full details are in the original sources.

Added 2 Nov 2026 · Updated 2 Nov 2026
Open case page ↗